A fullstack DevOps project for a demo radiology platform where medical imaging studies are received, anonymized, stored, analyzed with AI, and monitored through a complete containerized environment.
- Built a local Docker Compose stack with Orthanc, PostgreSQL, MinIO, FastAPI, AI inference, Prometheus, and Grafana
- Implemented a DICOM pipeline for metadata extraction, anonymization, preview generation, and upload to S3-compatible object storage
- Developed a FastAPI backend API and dashboard for studies, image viewing, AI results, audit events, and operational links
- Integrated a separate AI inference service with TorchXRayVision DenseNet121 for chest X-ray findings, confidence scores, and heatmap generation
- Stored AI results, metadata, and audit logs in PostgreSQL for traceability and history
- Built a demo workflow for radiographer upload and doctor review with automated AI processing
- Added Prometheus metrics and Grafana dashboards for both the API and AI service
- Implemented backup and restore flows for PostgreSQL, MinIO, and Orthanc
- Verified CI with GitHub Actions through real push, pull request, and merge workflows
- Deployed the platform across three Hetzner VPS nodes for app, data/imaging, and monitoring, using a private network and one Docker Compose setup per node